Periodic digital signals are commonly used in a variety of electronic devices. Probably the most common of periodic digital signals are clock signals that are typically used to establish the timing of a digital signal or the timing at which an operation is performed on a digital signal. For example, data signals are typically coupled to and from memory devices, such as synchronous dynamic random access memory (“SDRAM”) devices, in synchronism with a clock or data strobe signal.
As the speed of memory devices and other devices continue to increase, the “eye” or period in which a digital signal, such as a data signal, is valid becomes smaller and smaller, thus making the timing of a strobe signal or other clock signal used to capture the digital signal even more critical. In particular, as the size of the eye becomes smaller, the propagation delay of the strobe signal can be different from the propagation delay of the captured digital signal(s). As a result, the skew of the strobe signal relative to the digital signal can increase to the point where a transition of the strobe signal is no longer within the eye of the captured signal.
One technique that has been used to ensure the correct timing of a strobe signal relative to captured digital signals is to use a phase-lock loop (“PLL”) to generate the strobe signal. In particular, a phase-lock loop allows the timing of the strobe signal to be adjusted to minimize the phase error between the strobe signal and the valid eye of the digital signal. For example, as shown in FIG. 1, a conventional phase-lock loop 10 receives an input clock signal CLKIN and generates an output clock signal CLKOUT from the CLKIN signal. The phase-lock loop 10 includes a phase detector 12 that receives the input clock CLKIN signal and compares the phase of the CLKIN signal to the output clock signal CLKOUT. The phase detector 12 generates an error signal VE that is indicative of the phase error between the CLKIN signal and the CLKOUT signal. This error signal VE is applied to a loop amplifier 14, which normally has a relatively high gain. The loop amplifier 14 generates an amplified error signal VE+
Although the VE signal has a relatively low frequency component indicative of the phase error between the CLKIN and CLKOUT signals, it also normally includes harmonics of the CLKIN and CLKOUT signals. As explained below, these harmonics would cause the phase of the CLKOUT signal to periodically vary at a high frequency, which is a trait known as “phase noise.” To minimize the phase noise, the amplified VE signal is applied to a loop filter 16, which is normally a low-pass filter having a cutoff frequency that is well below the frequency of the CLKIN signal. The loop filter 16 therefore generates a relatively low frequency control signal VCON that is applied to a voltage controlled oscillator (“VCO”) 20. A single component, such as an operational amplifier (not shown), is often used for both the loop filter 16 and the loop amplifier 14. The VCO 20 generates the CLKOUT signal with a frequency that is proportional to the magnitude of the VCON signal.
In operation, the closed-loop nature of the phase-lock loop 10 causes the phase of the CLKOUT signal from the VCO 20 to be adjusted so that the phase of the CLKOUT signal differs from the phase of the CLKIN by a phase error that causes the VCON signal to have a magnitude that maintains the frequency of the CLKOUT signal equal to the frequency of the CLKIN signal. In general terms, a small phase error can be maintained by using a loop amplifier 14 having a larger gain since a given phase error will produce a larger control voltage VCON.
Another conventional phase-lock loop 30 is shown in FIG. 2. The phase-lock loop 30 is substantially identical in structure and operation to the phase-lock loop 10 of FIG. 1. Therefore, in the interest of brevity, identical components have been provided with the same reference numerals, and an explanation of their function and operation will not be repeated. The phase-lock loop 30 differs from the phase-lock loop 10 by including a frequency divider 34 in the signal path from the VCO 20 to the phase detector 12. The frequency divider 34 is programmable to reduce the frequency of the CLKOUT signal by dividing it by any integer value N. Therefore, if the CLKOUT signal has a frequency of F0, the signal fed back to the phase detector 12 will have a frequency of F0/N.
In operation, the closed loop nature of the phase-lock loop 30 will cause the VCON signal to have a value that ensures that the frequency of the signals applied to the phase detector 12 are equal to each other. Thus, if the CLKIN signal has a frequency of FIN, the frequency F0/N of the signal fed back to the phase detector 12 will also be FIN, i.e., F0/N=FIN. Solving this equation for F0, it can be seen that F0=N*FIN, i.e., the CLKOUT signal will have a frequency that is an integer multiple of the frequency of the CLKIN signal.
Although phase-lock loops have been successful in allowing digital signals to be captured in a digital device operating at a high speed, they are not without their disadvantages. In particular, phase-lock loops can consume a great deal of power, which can be a significant disadvantage in certain applications, such as in battery powered devices like laptop computers. The magnitude of the power consumed by phase-lock loops is a function of several parameters. In general, the power consumed by a phase-lock loop is directly proportional to the frequency of the signal generated by the loop since power is consumed each time a transistor is switched between two logic levels. Unfortunately, a high operating frequency is needed to match the high operating speed of digital devices, thus making it impractical to minimize power consumption. Also, a high operating frequency has the advantage of reducing the time required for the phase-lock loop to achieve a locked condition.
Phase-lock loops can also exhibit problem other than those related to power consumption. A clock signal produced by a phase-lock loop can have an unacceptable amount of phase noise, particularly if the loop amplifier 14 has a high gain, which, as explained above, is desirable to provide good phase control. While phase noise can be reduced by reducing the frequency response of the loop filter 16, doing so can reduce the ability of the loop to respond to variations in the frequency of the CLKIN signal and may unduly increase the time required for the loop to achieve lock.
The effect of phase noise and other noise sources can be explained with reference to the phase-lock loop shown in FIG. 3, which is the phase-lock loop 30 of FIG. 2 to which noise sources θN1, θN2, and θN3 have been added. Also shown in FIG. 3 are the gain of the phase detector 12 as KΦ, the transfer function of the loop amplifier 14 as ZF(S), and the transfer function of the VCO 20 as KVCO/S. The noise source θN1 is the phase noise in the CLKIN signal, which can result, for example, from variations in power supply voltage. The noise source θN2 is electrical noise in the loop filter 16, which can result, for example, from cross coupling of signals in the loop filter 16. The noise source θN3 is phase noise in the voltage controlled oscillator 20. The open loop gain G(S) of the phase-lock loop 30 is given by the formula G(S)=KΦZF(S)KVCO/S, and the transfer function between all of these noise sources and the output signal CLKOUT, can be expressed by the following formulae:HN1(S)=NG(S)/(1+G(S))  (Graph 1)HN2(S)=KVCO/S(1+G(S))  (Graph 2)HN3(S)=1/(1+G(S))  (Graph 3)
Graphs for these formulae are shown in FIG. 4. As explained below, similar graphs for an embodiment of the invention can be favorably compared to these graphs.
There is therefore a need for a phase-lock loop that can operate at a high frequency and yet consume relatively little power, and can operate over a wide frequency range and relatively quickly achieve a locked condition.